The difference between talking about AI and really using it

Every consultancy has an AI point of view. Every software vendor has an AI roadmap. Every conference agenda is full of agents, copilots and transformation. The problem is that too much of the conversation still lives in theory, and many still have no clear idea how to use AI well in practice.

At Bedford, we are more interested in what happens when AI meets a real Anaplan implementation: live stakeholders, imperfect requirements, tight delivery windows, contested decisions and business users who need the output to stand up in the room. That is where the useful lessons are emerging, and this is what we have learned from working there.

AI is already in the delivery room

On the implementations we are running, AI is not a future capability. It is already part of how the work gets done, on both sides of the table.

Business users are using tools such as Claude to review requirements/ Project teams are using it to summarise workshops and generate actions. Solution owners are asking itto challenge design decisions and find the gaps. Sponsors are using it to sense-check outputs and ask the questions they may not have thought to ask. And in some cases, customers are using AI to review the work we’re delivering.

We welcome that last one, because it reflects how we think the technology should be used. The goal is not to protect the implementation from scrutiny, but to make sure the implementation stands up to it.

AI doesn’t replace expertise. It increases the need for it.

AI can generate an answer very quickly, but what it cannot do is tell you whether that answer is right for your business.

We’ve watched AI produce model structures that look sensible but ignore who actually owns the plan. We’ve seen calculation logic that works technically but misses how the business really operates. We’ve seen recommendations that sound convincing until an experienced consultant spots the assumptions hidden underneath, or the source data being pulled through to reach a number that turns out to be wrong.

So our approach isn’t to let AI run unchecked. It’s to use it as a challenge function. Customers challenge our outputs, we challenge AI’s outputs, and together we arrive at something better. The result is usually a stronger solution, clearer documentation, and a more informed conversation.

The overlooked skill: reviewing the prompt, not just the answer

Most organisations focus on what the AI produced, but the smarter ones focus on what was asked.

A poorly written prompt can generate an impressive-looking answer that is fundamentally flawed. A well-constructed onecan surface real insight, identify risk, and accelerate a decision. That difference is becoming a genuine part of implementation governance.

We increasingly review AI-generated outputs alongside the prompts that created them, because the quality of the answer is directly linked to the quality of the question. This matters most exactly where AI is being used to validate model design, review user stories, or assess an implementation approach.The conversation is no longer simply “what did AI say?,” it’s “why did AI say it?”

AI as another layer of quality assurance

Implementation quality has traditionally relied on peer review, testing cycles, and governance processes. Those still matter but AI is starting to add another layer on top.

We have seen customers use it to review model documentation, challenge solution designs, test calculation logic, compare build outputs against requirements, and find inconsistencies across documentation before rollout. Most of the time this is not replacing existing quality processes. It is strengthening them. It becomes another reviewer in the room, one that never gets tired, never runs out of time, and can process a large volume of data almost instantly. The job the of experienced consultant is then to interpret and validate what it finds.

The same is true of project communication, where the impact is most immediate. Meeting capture, workshop summaries, action tracking and decision logs have always consumed a large share of project time.AI now handles much of that automatically. The value is not only the time saved. It is the visibility: decisions become easier to track, actions harder to lose, and a new stakeholder can catch up without reading weeks of historyOn a large Anaplan programme with several workstreams, that is a real advantage.

What we have learned

After seeing AI used by both delivery teams and customers, one thing is clear, the organisations gaining the most value are not treating AI as a replacement for expertise. Instead they are using it to amplify expertise: to ask sharper questions, challenge more assumptions, spot more risks and make better decisions.The technology accelerates the process, but people still provide the judgement.

So the real question is not whether AI can write a user story, propose a model structure, or summarise a workshop. It can.The question is whether the organisation has the expertise to validate, challenge, and govern what AI produces. That’s where implementation partners have to evolve.

At Bedford, we don’t just talk about AI-enabled delivery.We’re already working alongside customers who are embedding AI into their implementation programmes, using it to review our work, challenge assumptions, capture actions, improve quality assurance, and accelerate decision-making, and we use AI ourselves across scoping, modelling, forecasting, training, governance and handover . The future of implementation isn’t humans or AI; it’s the combination of the two. The organisations that get that balance right will move faster, make better decisions, and extract more value from their Anaplan investment than those still debating what AI might do next.

What to read and do next

A related read.

If this piece is about using AI well on the way in, its companion is about the environment it lands in: AI will not fix weak planning foundations. It will expose them. Together they make the same point from two directions — the value is in the expertise and the environment, not the tool.

Let’s talk. If this feels recognisable, we would value the conversation. Reach us at info@bedfordconsulting.com or follow Bedford Consulting on LinkedIn.